Here are some design decisions about which there are conflicting goals.
Many, probably most, leaf nodes will themselves be multiprogrammed with several concurrent activities, each with network business unrelated to other network business at the same node. It seems almost mandatory to use additional turn op data to route incoming packets to the particular activity. This fits well with network backpressure schemes.
Today (2012 Sept 22) I see no way to enforce the 0 turn op convention. Nor do I see a need to. It is a useful convention to address the switching logic of some intermediate node. There may be some conventional things you might want to say to such nodes but we have not described any yet.
Today (2012 Sept 28) it seems obvious that a zero length destination string is the obvious indication that the new holder is the addressee. How could be otherwise?
(2012 Nov 29) The stuff above assumes that turn ops are interface numbers. The same might have been said for interface numbers in place of turn ops when we adopt the XOR scheme.